Identification of key regulators in pancreatic ductal adenocarcinoma using network theoretical approach

Pancreatic ductal adenocarcinoma (PDAC) is a common type of cancer originating from the pancreatic glands and is characterized by a rapidly progressive course and a dismal prognosis. Combination strategies targeting multiple signaling pathways supporting tumor growth and propagation are active areas of contemporary research that will likely transform the treatment paradigm of pancreatic cancer. PDAC has poor clinical outcomes due to delayed detection, chemotherapy resistance, and absence of specific targeted therapies. Identifying novel therapeutic targets and early biomarkers may improve PDAC management. A connectivity Map can provide information on signaling pathways activated by drugs and assist in development. Exploiting aberrant metabolic processes in PDAC cells is a promising strategy. The incidence of PDAC is rising and it is a leading cause of cancer-related death. Targeted therapy has had limited success due to molecular heterogeneity and complex tumor microenvironment (TME). Retinoids, predictive molecular markers, and gene panel testing may aid personalized therapies. Accurate diagnosis is crucial to distinguish PDAC from other pancreatic neoplasms. Novel therapeutic targets include genomic alterations, TME, and metabolism; sequencing enables personalized therapies; immunotherapy shows promise in specific PDAC subsets (dMMR/MSI-H). Dense fibrotic stroma hinders drug delivery and promotes resistance. Advances in surgery, immunotherapy, and targeted therapies are being explored. The TME plays a crucial role in PDAC development, progression, and resistance; immunotherapies and targeted approaches focusing on specific pathways and TME components (e.g., fibroblasts) are under study. Standard-of-care chemotherapy (gemcitabine/nab-paclitaxel, FOLFIRINOX) improves outcomes but the 5-year survival remains low. Adjuvant chemotherapy offers survival advantage; chemoradiation addition is not recommended. Understanding PDAC risk factors supports screening and lifestyle counseling. A systemic approach leveraging network medicine is essential, considering interactions among genes rather than single-gene effects. This study analyzes PDAC protein-protein interaction networks constructed from differentially expressed genes to understand architectural principles and predict key regulators, aiming to advance understanding of PDAC initiation and progression and strengthen therapeutic approaches. The study establishes an unbiased catalog of up- and down-regulated gene expression from 12 PDAC samples, identifying 33 prominent genes implicated in cytoskeleton rearrangement, tissue development, and immune system activation.

The paper reviews current PDAC management and research: Standard chemotherapy regimens (gemcitabine/nab-paclitaxel, FOLFIRINOX) improve outcomes but survival remains poor; adjuvant chemotherapy benefits resected PDAC, while chemoradiation does not. Molecular heterogeneity and dense stromal TME contribute to treatment resistance and immunotherapy challenges. Emerging approaches include targeted therapies against specific signaling pathways, TME modulation (fibroblasts and CAFs), immunotherapy (checkpoint inhibitors in dMMR/MSI-H tumors), and exploiting altered cancer metabolism. Advances in genome sequencing foster personalized therapy. Diagnostic challenges include distinguishing PDAC from ACC and PNET. Network medicine is emphasized to understand complex disease modules. Prior studies highlight roles of retinoids, predictive molecular markers, and gene panels; the physical barrier of fibrotic stroma impairs drug delivery. The importance of analyzing gene/protein interaction networks for biomarker and target discovery is underscored.

Data were obtained from the RNA-seq dataset GSE171485 (NCBI GEO). Analyses were conducted in R/RStudio (4.2.2). Differentially expressed genes (DEGs) were identified using fold change (log2 transformation) and LIMMA, applying linear modeling and empirical Bayes to perform t- and F-tests, with Benjamini-Hochberg adjusted p-value ≤ 0.05 and significance p ≤ 0.05; threshold log2 |FC| ≥ 1. Visualization used ggplot2. A PDAC protein-protein interaction (PPI) network was constructed with STRING (interaction score > 0.4) incorporating physical/functional associations from literature text-mining, co-expression, genomic context, predictions, experiments, and aggregated databases. Network visualization/analysis used Cytoscape (v3.6.1). Gene ontology (GO) enrichment of DEGs (Molecular Function, Biological Process, Cellular Component) employed DAVID. Global network topological properties were delineated: degree distribution P(k), clustering coefficient C(k), neighborhood connectivity C_N(k), and centralities, with power-law fits verified via Clauset et al. statistical fitting (2500 random samplings). Assortativity was assessed. Community detection used the leading eigenvector (LEV) method with igraph to identify modules, sub-modules, and motifs (G(3,3)); a sub-module was considered a community if it contained at least one motif. miRNA targeting of key genes was analyzed with MIENTURNET relying on TargetScan predictions; significant enrichments noted. Drug-gene interactions were queried in DGIdb for approved drugs, capturing interaction types (inhibition, activation, binding).

• DEG analysis identified 772 DEGs: 431 up-regulated and 341 down-regulated (log2 |FC| ≥ 1, p < 0.05, Padj ≤ 0.05).
• Down-regulated GO BP terms included potassium ion transport, cell-cell signaling, neuropeptide signaling pathway, muscle contraction, cellular response to zinc ion, response to hypoxia, negative regulation of cytosolic calcium ion concentration, positive regulation of apoptotic process. CC terms included extracellular region, mitochondrion, mitochondrial inner membrane, endoplasmic reticulum, potassium channel complex. MF terms included protein binding, methyltransferase activity, structural constituent of ribosome, neuropeptide hormone activity. KEGG pathways enriched: metabolic pathways, peroxisome, cAMP signaling, chemical carcinogenesis – DNA adducts/reactive oxygen species, biosynthesis of cofactors, ribosome, mineral absorption.
• Up-regulated GO BP terms included cell adhesion, cell-cell adhesion, positive regulation of cell migration, extracellular matrix organization, tissue development, epithelial cell differentiation, integrin-mediated signaling pathway, response to hypoxia. MF terms included cadherin binding, extracellular matrix structural constituent, virus receptor activity, calcium ion binding, insulin-like growth factor I binding, integrin binding, calcium-dependent protein binding. CC terms included plasma membrane, extracellular exosome, perinuclear region of cytoplasm, adherens junction, extracellular space. KEGG pathways enriched: ECM-receptor interaction, p53 signaling pathway, pathways in cancer, proteoglycans in cancer, focal adhesion, tight/adherens junction, PI3K-Akt signaling.
• Network topology: Power-law behavior confirmed (Clauset method; p-values > 0.1 vs 2500 random samplings; goodness-of-fit ≤ 0.35). Exponents: α ≈ 0.294 (P(k)), β ≈ 0.165 (C(k)), γ ≈ 0.122 (C_N(k)); centrality-vs-degree exponents: γ0 ≈ 0.302 (betweenness), δ0 ≈ 0.095 (closeness), ψ0 ≈ 1.008 (eigenvector). Positive assortativity (φ > 0) indicates assortative mixing. Interpretation: weak hierarchical, scale-free fractal network with hubs sparsely distributed; high-degree nodes less clustered (β < 1), hierarchy of hubs.
• Community detection (LEV) revealed organization across five hierarchical levels; tracing G(3,3) motifs from top to motif level identified 33 key regulators present at all levels: F10, PLAT, F2R, MMP1, CTSG, F2RL1, PDGFA, FGF13, PTN, APOC2, SDC1, SDC4, AP2B1, AP1S3, DNM2, EGFL7, RASIP1, SOX18, CLDN2, CLDN7, CDH3, CLDN5, TJP1, PCDH1, RNF213, EPSTI1, XAF1, VIP, GAL, GNRH1, NMU, VIPR2, GNG11.
• Low-degree nodes can be essential regulators due to topological position: F2RL1 (degree 4), F10 (4), APOC2 (8), CLDN2 (11), PCDH1 (3) formed motifs with higher-degree partners (e.g., F2RL1 with MMP1 degree 36; F10 with PLAT degree 17; APOC2 with SDC1 degree 40; PCDH1 with TJP1 degree 44). Sample topological metrics (Table 2) illustrate influence of low-degree bridge nodes.
• Enrichment of key regulators: Molecular functions—G-protein beta-subunit binding (F2R, F2RL1, GNG11; p ≈ 0.000697; FE ≈ 74.1), receptor binding (EGFL7, F2R, NMU, F2RL1, PLAT; p ≈ 0.00433), serine-type endopeptidase activity (F10, MMP1, CTSG, PLAT; p ≈ 0.00346), growth factor activity (PDGFA, PTN, FGF13; p ≈ 0.0406). Cellular components—extracellular region and plasma membrane enriched. Biological processes—positive regulation of cell proliferation; positive regulation of protein kinase B (AKT) signaling; calcium-independent cell-cell adhesion via plasma membrane adhesion molecules. KEGG—neuroactive ligand-receptor interaction (VIPR2, GAL, F2R, NMU, F2RL1, GNRH1, CTSG, VIP); Cell adhesion molecules (CLDN5, CDH3, SDC4, CLDN7, SDC1, CLDN2); Leukocyte transendothelial migration (CLDN5, CLDN7, CLDN2).
• Drug–gene interactions (DGIdb): 91 interactions across 14 genes and 79 approved drugs; highest interaction counts for MMP1, F10, F2R, GNRH1 (>20 drugs). Examples: F10—multiple anticoagulants (e.g., apixaban, rivaroxaban, edoxaban); PLAT—urokinase, aminocaproic acid; F2R—vorapaxar, argatroban; MMP1—marimastat, doxycycline; SDC1—heparin, indatuximab ravtansine; EGFL7—parsatuzumab; CDH3—PF-06671008; TJP1—risperidone, genistein; VIP—digoxin, lisinopril; GAL—hydrocortisone, mirtazapine; GNRH1—goserelin, leuprolide.
• Transcription factors: 15 key TFs regulating the key genes (e.g., NF1, SP1, SP3, TWIST1, MYB, GATA3, POU2F1, JUN, PPARG, HDAC1, STAT1, CREB1, STAT3, RELA, NFKB1) with significant p-values.
• miRNAs: 7 predicted key miRNAs (e.g., hsa-miR-335-5p targeting 9 genes; hsa-miR-1-3p; hsa-miR-526b-3p; hsa-miR-615-3p; hsa-miR-106b-5p; hsa-miR-17-5p; hsa-miR-124-3p) indicating regulatory layers affecting key genes and TFs.

Topological analysis of the PDAC PPI network (605 nodes, 2698 edges) indicates a weak hierarchical, scale-free fractal structure supporting efficient information processing. Using Girvan–Newman/LEV methods, the network exhibits local clustering of low-degree nodes into communities coupled with global hub connectivity, reflecting self-similar organization that contributes to stability. Integrating differential expression with network topology enabled identification of 33 key regulators deeply embedded across hierarchical levels, consistent with their role as functional bottlenecks. Several identified regulators (e.g., F10, PLAT, F2R/PAR2, MMP1, PDGFA, FGF13, PTN, TJP1, CLDN family, CDH3, PCDH1, RNF213, EPSTI1, VIP/VIPR2, NMU, GNRH1, GNG11) have literature support for involvement in cancer progression, invasion, metastasis, EMT, angiogenesis, and therapy resistance. The prominence of neuroactive ligand–receptor interaction and leukocyte transendothelial migration pathways ties PDAC aggressiveness to nerve-associated invasion, immunosuppressive microenvironment, pain signaling, and tumor-promoting inflammation. Low-degree bridge nodes (e.g., F2RL1, F10, APOC2, CLDN2, PCDH1) demonstrate that essentiality depends on topological position, not degree alone. Drug–gene interaction mapping highlights actionable targets (e.g., anticoagulants for F10, protease inhibitors for F2R, MMP inhibitors for MMP1), suggesting repositioning opportunities; TFs (e.g., NFKB1, STAT1/3, CREB1) and miRNAs (e.g., miR-335-5p) provide regulatory intervention points. These findings support biomarker discovery and precision therapy development in PDAC, warranting experimental validation in larger cohorts.

A comprehensive RNA-seq analysis comparing PDAC to healthy controls identified DEGs and enriched biological functions and pathways. Network topology analysis revealed a weak hierarchical, scale-free fractal organization and enabled discovery of 33 key regulators implicated in PDAC progression. Approved drugs, transcription factors, and miRNAs targeting these regulators were cataloged, offering avenues for therapeutic targeting and biomarker development. The study underscores the utility of integrated transcriptomic and network analyses for personalized treatment planning in PDAC and calls for experimental validation to confirm clinical relevance.

The study is limited by a constrained sample size and lack of experimental validation. Further investigation is required to verify the expression and function of the identified key regulators in PDAC and to assess generalizability across larger, independent cohorts.